| name | botnet |
| description | Master skill for managing the owner-operated Botnet Mac fleet across gist, botnet, and kokoro-coreml. Use when checking fleet health, rolling Mac workers, debugging scrape.remote_browser, tts.kokoro_coreml, or precompute workers, handling LaunchAgents, SSH fan-out, Cloudflare control-plane health, Kokoro/Core ML runtime issues, botnet fleet env vars, or writing fleet debugging notes. |
Botnet Fleet
Purpose
Use this as the master skill for Botnet fleet operations. It coordinates the
cross-repo system made of:
botnet: control plane, SSH rollout scripts, remote browser worker, native
Kokoro worker, precompute workers, fleet inventory, and operator notes.
gist: upstream extension/runtime behavior and Gist-facing product contract.
kokoro-coreml: Swift/Core ML model export, runtime assets, profiling, and
validation evidence for native TTS.
Read reference.md before acting on anything non-trivial. It is
the path catalog for related skills, runbooks, notes, environment variables, and
historical decisions.
Use When
- The user mentions the Botnet fleet, Mac fleet,
m2-studio, irvine-m1,
m2-air, SSH rollout, update-fleet, update-host, or fleet health.
- The task involves
scrape.remote_browser, tts.kokoro_coreml, reader-audio
precompute workers, worker LaunchAgents, queue depth, worker freshness,
workersByMachine, active slots, or per-machine capability readiness.
- The failure crosses
gist, botnet, and kokoro-coreml, or depends on which
repo owns the broken contract.
- The task asks where Botnet fleet env vars, worker keys, admin bearer keys, or
Cloudflare deploy credentials live.
First Rule
Start in botnet unless there is strong evidence the root cause belongs
elsewhere. botnet owns the live control plane and fleet scripts; gist and
kokoro-coreml are upstream or dependency repos for specific contracts.
Required Workflow
- Read
reference.md, then load only the related skill or note
for the current failure class.
- Prove current fleet state before changing anything:
cd /Users/mattmireles/Documents/Github/botnet and run the checked-in
health workflow from fleet-health-check when live status matters. From this
operator Mac you can SSH directly to m2-studio and irvine-m1 for remote
inspection and rollouts—see reference.md § Operator Workstation.
- Repair the failing primitive, not the whole system:
control-plane health, SSH reachability, LaunchAgent state, scraper worker,
Kokoro worker, precompute worker, or Cloudflare deploy.
- Use
fleet-update for rollouts, fleet-health-check for health triage,
remote-worker for browser worker incidents, and Kokoro/Core ML skills or
docs for TTS runtime defects.
- Verify with the operator command that matches the repair. Do not call the
fleet healthy from a single local symptom.
- For debugging that takes real investigation, write or update notes before
finishing. Prefer
botnet/README/notes/ for fleet incidents; use gist or
kokoro-coreml notes only when that repo owns the durable lesson.
Guardrails
- Do not print secret values from
.env, worker keys, admin bearer keys, or
Cloudflare tokens.
- Do not deploy the shared Cloudflare API worker unless the user asked for a
deploy or the rollout requires
pnpm update-fleet --deploy-api.
- Do not rewrite imported Gist worker code inside
botnet casually. Preserve
upstream import boundaries and sync deliberately.
- Do not treat
gist's older Mac mini runbooks as the live source of truth when
botnet has newer Botnet-branded fleet docs.
- Do not change Kokoro model/runtime policy without checking
kokoro-coreml
evidence and botnet Kokoro throughput notes.